Pricing
Silicon Network and public cloud list prices
The rate of each model on Silicon Network next to the on-demand list price public clouds publish for the same open-weight model, per 1M tokens in USD. The cloud prices were read on each cloud's own pricing page on October 29, 2026. They change: check the source before relying on one.
What is compared
Silicon Network rates are the published price list, read when this page was last built, and bill the same for every request. Silicon Network serves the quantized weights each model page lists, on Apple Silicon Macs; a cloud may serve the same model at another precision. Batch, cached input, provisioned throughput and regional surcharges are left out on both sides. A model a cloud does not list for on-demand use has no row for that cloud.
Llama 3.2 1B
| Provider | Offered as | Region | Input | Output |
|---|---|---|---|---|
| Silicon Network | llama-3.2-1b (4bit) | Wherever a Mac serves it | $0.50 * | $1.50 * |
| Amazon Web Services | Amazon Bedrock: Llama 3.2 Instruct (1B) | US East (N. Virginia) | $0.10 | $0.10 |
* This model has no rate of its own on the price list and bills at the network default.
- Amazon Web Services, read October 29, 2026 on aws.amazon.com: “Llama 3.2 Instruct (1B): 0.0001 input, 0.0001 output per 1K tokens. The page fills its US prices in from AWS's price file (b0.p.awsstatic.com, bedrock.json), which states them per 1K tokens; multiplied by 1,000.”
Llama 3.2 3B
| Provider | Offered as | Region | Input | Output |
|---|---|---|---|---|
| Silicon Network | llama-3.2-3b (4bit) | Wherever a Mac serves it | $0.50 * | $1.50 * |
| Amazon Web Services | Amazon Bedrock: Llama 3.2 Instruct (3B) | US East (N. Virginia) | $0.15 | $0.15 |
* This model has no rate of its own on the price list and bills at the network default.
- Amazon Web Services, read October 29, 2026 on aws.amazon.com: “Llama 3.2 Instruct (3B): 0.00015 input, 0.00015 output per 1K tokens. The page fills its US prices in from AWS's price file (b0.p.awsstatic.com, bedrock.json), which states them per 1K tokens; multiplied by 1,000.”
Llama 3.1 8B
| Provider | Offered as | Region | Input | Output |
|---|---|---|---|---|
| Silicon Network | llama-3.1-8b (4bit) | Wherever a Mac serves it | $0.50 * | $1.50 * |
| Amazon Web Services | Amazon Bedrock: Llama 3.1 Instruct (8B) | US East (N. Virginia) | $0.22 | $0.22 |
* This model has no rate of its own on the price list and bills at the network default.
- Amazon Web Services, read October 29, 2026 on aws.amazon.com: “Llama 3.1 Instruct (8B): 0.00022 input, 0.00022 output per 1K tokens. The page fills its US prices in from AWS's price file (b0.p.awsstatic.com, bedrock.json), which states them per 1K tokens; multiplied by 1,000.”
Llama 3.1 70B
| Provider | Offered as | Region | Input | Output |
|---|---|---|---|---|
| Silicon Network | llama-3.1-70b (4bit) | Wherever a Mac serves it | $0.50 * | $1.50 * |
| Amazon Web Services | Amazon Bedrock: Llama 3.1 Instruct (70B) | US East (N. Virginia) | $0.72 | $0.72 |
* This model has no rate of its own on the price list and bills at the network default.
- Amazon Web Services, read October 29, 2026 on aws.amazon.com: “Llama 3.1 Instruct (70B): 0.00072 input, 0.00072 output per 1K tokens, standard (not latency optimized). The page fills its US prices in from AWS's price file (b0.p.awsstatic.com, bedrock.json), which states them per 1K tokens; multiplied by 1,000.”
Llama 3.3 70B
| Provider | Offered as | Region | Input | Output |
|---|---|---|---|---|
| Silicon Network | llama-3.3-70b (4bit) | Wherever a Mac serves it | $0.50 * | $1.50 * |
| Amazon Web Services | Amazon Bedrock: Llama 3.3 Instruct (70B) | US East (N. Virginia) | $0.72 | $0.72 |
| Google Cloud | Vertex AI Model Garden: Llama 3.3 70B | Not stated on the page | $0.72 | $0.72 |
| Microsoft Azure | Azure AI Foundry Models: Llama 3.3 70B (Global) | Global deployment, East US price | $0.71 | $0.71 |
* This model has no rate of its own on the price list and bills at the network default.
- Amazon Web Services, read October 29, 2026 on aws.amazon.com: “Llama 3.3 Instruct (70B): 0.00072 input, 0.00072 output per 1K tokens. The page fills its US prices in from AWS's price file (b0.p.awsstatic.com, bedrock.json), which states them per 1K tokens; multiplied by 1,000.”
- Google Cloud, read October 29, 2026 on cloud.google.com: “Meta's Llama models, price per 1M (USD): Llama 3.3 70B, input $0.72, output $0.72.”
- Microsoft Azure, read October 29, 2026 on azure.microsoft.com: “Llama 3.3 70B Global: input 0.71, output 0.71 per 1M tokens (Data zone and regional deployments 0.781).”
DeepSeek V3.1
| Provider | Offered as | Region | Input | Output |
|---|---|---|---|---|
| Silicon Network | deepseek-v3.1 (4bit) | Wherever a Mac serves it | $0.50 * | $1.50 * |
| Amazon Web Services | Amazon Bedrock: DeepSeek-V3.1 | US West (Oregon); not offered in US East (N. Virginia) | $0.58 | $1.68 |
| Google Cloud | Vertex AI Model Garden: DeepSeek-V3.1 | Not stated on the page | $0.60 | $1.70 |
| Microsoft Azure | Azure AI Foundry Models: DeepSeek-V3.1 | East US | $1.23 | $4.94 |
* This model has no rate of its own on the price list and bills at the network default.
- Amazon Web Services, read October 29, 2026 on aws.amazon.com: “Standard tier, DeepSeek-V3.1: 0.00058 input, 0.00168 output per 1K tokens. The page fills its US prices in from AWS's price file (b0.p.awsstatic.com, bedrock.json), which states them per 1K tokens; multiplied by 1,000.”
- Google Cloud, read October 29, 2026 on cloud.google.com: “DeepSeek's models, price per 1M (USD): DeepSeek-V3.1, input $0.60, output $1.70.”
- Microsoft Azure, read October 29, 2026 on azure.microsoft.com: “DeepSeek-V3.1: input 1.23, output 4.94 per 1M tokens.”
Qwen3 235B
| Provider | Offered as | Region | Input | Output |
|---|---|---|---|---|
| Silicon Network | qwen3-235b (4bit) | Wherever a Mac serves it | $0.50 * | $1.50 * |
| Amazon Web Services | Amazon Bedrock: Qwen3 235B A22B 2507 | US West (Oregon); not offered in US East (N. Virginia) | $0.22 | $0.88 |
| Google Cloud | Vertex AI Model Garden: Qwen3-235B-A22B-Instruct-2507 | Not stated on the page | $0.22 | $0.88 |
* This model has no rate of its own on the price list and bills at the network default.
- Amazon Web Services, read October 29, 2026 on aws.amazon.com: “Standard tier, Qwen3 235B A22B 2507: 0.00022 input, 0.00088 output per 1K tokens. The page fills its US prices in from AWS's price file (b0.p.awsstatic.com, bedrock.json), which states them per 1K tokens; multiplied by 1,000.”
- Google Cloud, read October 29, 2026 on cloud.google.com: “Qwen's models, price per 1M (USD): Qwen3-235B-A22B-Instruct-2507, input $0.22, output $0.88.”
GPT-OSS 20B
| Provider | Offered as | Region | Input | Output |
|---|---|---|---|---|
| Silicon Network | gpt-oss-20b (MXFP4-Q8) | Wherever a Mac serves it | $0.50 * | $1.50 * |
| Amazon Web Services | Amazon Bedrock: gpt-oss-20b | US East (N. Virginia) | $0.07 | $0.30 |
| Google Cloud | Vertex AI Model Garden: gpt-oss-20b | Not stated on the page | $0.07 | $0.25 |
* This model has no rate of its own on the price list and bills at the network default.
- Amazon Web Services, read October 29, 2026 on aws.amazon.com: “Standard tier, gpt-oss-20b: 0.00007 input, 0.0003 output per 1K tokens. The page fills its US prices in from AWS's price file (b0.p.awsstatic.com, bedrock.json), which states them per 1K tokens; multiplied by 1,000.”
- Google Cloud, read October 29, 2026 on cloud.google.com: “OpenAI's models, price per 1M (USD): gpt-oss-20b, input $0.07, output $0.25.”
GPT-OSS 120B
| Provider | Offered as | Region | Input | Output |
|---|---|---|---|---|
| Silicon Network | gpt-oss-120b (MXFP4-Q8) | Wherever a Mac serves it | $0.50 * | $1.50 * |
| Amazon Web Services | Amazon Bedrock: gpt-oss-120b | US East (N. Virginia) | $0.15 | $0.60 |
| Google Cloud | Vertex AI Model Garden: gpt-oss-120b | Not stated on the page | $0.09 | $0.36 |
| Microsoft Azure | Azure AI Foundry Models: gpt-oss-120b | East US | $0.15 | $0.60 |
* This model has no rate of its own on the price list and bills at the network default.
- Amazon Web Services, read October 29, 2026 on aws.amazon.com: “Standard tier, gpt-oss-120b: 0.00015 input, 0.0006 output per 1K tokens. The page fills its US prices in from AWS's price file (b0.p.awsstatic.com, bedrock.json), which states them per 1K tokens; multiplied by 1,000.”
- Google Cloud, read October 29, 2026 on cloud.google.com: “OpenAI's models, price per 1M (USD): gpt-oss-120b, input $0.09, output $0.36.”
- Microsoft Azure, read October 29, 2026 on azure.microsoft.com: “Open source models, pricing per 1M tokens: gpt-oss-120b, input 0.15, output 0.60 (us-east).”
Not listed
Models in the Silicon Network catalog that none of these clouds lists for on-demand, per-token use have no comparison here. Every model and its rate is on the model pages and the pricing page.